This project is about testing digital integrated circuits during their manufacture -- production testing. Production testing attempts to identify manufactured parts that do not perform according to their specification (defective parts) as well as parts that will fail early in their operating life (weak parts). Sometimes parts are also sorted by their maximum operating speed (binning). For some selected parts, the exact causes of their deviation from expected functionality must be identified (diagnosed) in order to improve the manufacturing process. Improved IC production test techniques are necessary because current techniques are too expensive and are projected to become increasingly unsatisfactory for future IC designs. The manufacturing cost per transistor is decreasing for newer technologies while the cost per transisitor for production testing is remaining approximately constant. In other words, the fraction of the total cost of manufacturing process due to testing is increasing. There is interest in reversing this trend.
Specific research goals are to develop test techniques to replace current burn-in, delay test and the single-stuck fault coverage metric. Other goals will to be compare test results for different generations of technology to find out to what extent results from one technology carry forward to a next generation, and to compare the effectiveness of functional tests with structured tests (for example, sequential circuit test patterns versus scan patterns). Experimental data obtained from ATE will be analyzed. This data will be used to suggest new techniques and to determine their effectiveness. We will use two types of experimental data: some is industrial data that we obtain as part of a joint study with a company; the other is data that we collect ourselves.
